论文部分内容阅读
在医学研究中,常常使用受试者操作特性曲线(ROC)曲线来研究两样本的比较问题.Lloyd~([1])构造了ROC曲线的核平滑估计,并给出了其渐近偏差以及渐近标准差.此外,当还可以获悉某一处理组上的辅助信息时,Zhou,Zhou&Ma~([2])利用经验似然的方法构造了ROC曲线的核平滑经验似然估计.本文利用“亏量”这个概念比较了带有辅助信息的情况下,对核平滑经验似然估计与完全经验似然估计进行了比较.并给出了核平滑经验似然估计优于完全经验似然估计的结论,并且随着样本容量的增大,该亏量也是无限增大的.“,”Receiver operating characteristic(ROC)curve is often used to study and compare two-sample problems in medicine.Lloyd~([1]) has proposed estimating the ROC curve from kernel smoothing techniques and has presented asymptotic forulas for the bias and standard deviation of the resulting curve estimator.When more information may be available on one treatment,Zhou,Zhou & Ma~([2]) proposed a smoothed empirical likelihood estimator for ROC curve. In this article,we compare the asymptotic accuracy of the kernel smoothed empirical likelihood estimator with the fully empirical likelihood estimator in the presence of auxiliary information using the concept of deficiency.It is shown that the fully empirical likelihood estimator is deficient compared to the kernel smoothed empirical likelihood estimator,and that this deficiency is unbounded as sample size increases.